Abstract: The brain is not only a remarkable computational organ - capable of feats that stymie the best computers and robots - it is the generator of our thoughts and actions. Yet modern systems neuroscience has principally asked how the brain transforms inputs into outputs. This approach has deep historical roots: both Descartes and Sherrington saw the nervous system as a massively elaborated reflex. The approach also produced critical early successes: the descriptions by Mountcastle, Hubel, and Wiesel, of how sensory stimuli drive single- neuron responses. Yet the brain is clearly more than a glorified input-output device. The neural networks within it do not just respond reflexively to external stimuli, they also generate their ow activity. In doing so they produce thoughts, plans, decisions and actions. As the study of such processes becomes increasingly central to systems neuroscience, we will need to become increasingly concerned with internal neural dynamics: how neural circuitry shapes and generates the responses that allow us to act upon the world. We will become less interested in how individual neurons reflect external stimuli. We will become much more interested in the dynamics of how neural activity sustains and shapes itself over time. I believe this rising interes in internal neural dynamics will drive large changes in the conceptual, analytical, and experimental paradigms employed by systems neuroscience. The first changes will focus on collecting, visualizing, and analyzing data that can reveal underlying dynamics: how the state of the neural circuit at one point in time leads lawfully to the state of the neural circuit at the net point in time. The focus will then shift to designing experiments that most effectively probe dynamics. Such experiments will borrow techniques from the physical sciences and from engineering, but will initially be based on the traditional behavioral paradigm of systems neuroscience in which animals are trained to produce tightly-controlled behavior. However, I believe the traditional experimental framework will give way to a new one. Instead of indirectly influencing neural activity by operantly conditioning behavior, we will directly monitor and operantly condition the internally generated neural activity itself. This methodology will be built upon the technical platform recently developed in the service of neuro-motor prostheses, but will serve a basic scientific purpose: it will give the experimenter unprecedented control over the system they are trying to understand, and allow stringent tests of hypotheses regarding dynamics. My goal is to help build this emerging paradigm. A subsequent but equal goal is to leverage our growing understanding of neural dynamics. I believe that we should be able to develop a new class of neural prosthetic device that uses the dynamic patterns of motor cortex activity to drive artificial locomotion. I believe this is both the best way to demonstrate that ou hard-won knowledge of dynamics is meaningful, and that it may be one of the most effective ways to develop a neuro-motor prosthesis that will help significant numbers of people.